On the sample variance of explosive random coefficient autoregressive processes

This work investigates the behavior of the sample variance of an explosive random coefficient autoregressive process y t = ( a + u t ) y t − 1 + ε t . It is shown that the simulated sample variance has a distribution when a 2 < 1 and a 2 + σ u 2 = 1 . Moreover, the variance of y t when a = − 1 is found to be three times larger as compared with the case where a = 1 .